Research written by
engineers, for engineers
who read past the abstract.
Original technical research on blockchain infrastructure, AI engineering, cloud security, and enterprise architecture. Every paper published here is based on systems we've built, problems we've encountered, and data we've measured - not survey responses and market projections.
All papers, sorted by recency.
Every paper is free to download, free to cite, and free to share. No gates, no registration, no drip sequences.
Smart Contract Security: A Practitioner's Framework for DeFi Protocol Risk
Economic attack modelling, oracle design, and the 11 vulnerability classes that account for 89% of DeFi exploits by value lost. Covers flash loan attack taxonomy, governance attack resistance, and a comparative analysis of audit tooling - from Slither to Echidna to manual review. Based on 180 contract audits and three real-world post-mortems.
Production LLM Architecture: Latency, Cost, and the Reliability Triangle
A practical guide to deploying large language models in production - prompt caching strategies, model routing based on task complexity, fallback chains, and cost attribution per business unit. Includes benchmarks across GPT-4o, Claude 3.5 Sonnet, and Gemini 1.5 Pro on 12 enterprise task categories, with p50 and p99 latency data from our production deployments.
Three multi-part series on topics
that need more than one paper.
Some subjects require a sequence of connected papers to cover properly. These series are designed to be read in order, though each paper stands alone.
From tokenomics to post-deployment monitoring - a complete curriculum for engineers building protocols that hold up under adversarial conditions. Each paper focuses on one phase of the security lifecycle.
How to build, operate, and evolve a Kubernetes-based internal developer platform - from initial cluster design through to a mature platform engineering function with CoE governance.
Moving AI from proof-of-concept to production - the engineering decisions, cost management strategies, and evaluation frameworks that determine whether LLM features remain reliable under real load.
New research published
quarterly. No filler between.
We publish one to three substantial papers per quarter - never more than we can back with real engineering work. Sign up and we'll email you when something new is ready. One email per paper, no marketing, unsubscribe any time.